Device-to-device (D2D) multicast communication is considered as a potential solution to improve the spectral efficiency of cellular networks. This paper focuses on resource allocation in underlay D2D multicast networks where multiple D2D multicast groups (MGs) share the uplink frequency channels with multiple cellular users (CUs). An optimization problem that maximizes the system throughput while fulfilling the maximum power constraint of every mobile user and ensuring a certain level of quality of service (QoS) to every CU and D2D multicast group is formulated. This formulation leads to mixed integer non-linear programming (MINLP), which is computationally intractable for large scale networks. Therefore, to find a feasible solution, we propose a channel sharing algorithm which determines how many MGs can share a channel with certain QoS guarantees to CU and D2D MGs. Then, we propose a power allocation algorithm that maximizes the system throughput while satisfying the various constraints. The impact of geographical spread of MGs, number of MGs, maximum available transmission power, and QoS requirements of every CU on achievable system throughput is analyzed. Numerical results show the efficacy of proposed model in terms of D2D MG's throughput and spectrum efficiency.